Building on a recent study on scientific collaboration networks, we propose an induced diffusion percolation model that brings superactive nodes into focus. Defined as active nodes surrounded by at least k active or superactive neighbors, superactive nodes play a key role in innovation diffusion by inducing their neighbors to adopt an innovation. We investigate the induced diffusion percolation model using the modified Newman–Ziff algorithm on two-dimensional lattices (square and triangular lattices) and regular random networks with and without clustering. The induction by superactive nodes leads to a first-order percolation phase transition in two-dimensional lattices and a double transition – a continuous percolation transition followed by a discontinuous jump of the order parameter of the largest cluster’s strength – in regular random networks. Whereas clustering increases the percolation threshold in the classical percolation model on regular random networks, it decreases the critical initial activation probability that triggers a discontinuous jump of the induced activation.